An algorithm that generalizes to novel noise segments to improve speech intelligibility for hearing-impaired listeners

Abstract

Machine-learning algorithms to extract speech from background noise hold considerable promise for alleviating limitations associated with hearing impairment. One of the most important considerations for implementing these algorithms into devices such as hearing aids and cochlear implants involves their ability to generalize to conditions not employed during the training stage. A major challenge involves the generalization to novel noise segments. In the current study, sentences were extracted from multi-talker babble and from cafeteria noise using an algorithm that estimates the ideal ratio mask by employing deep neural networks. Importantly, the algorithm was trained on segments of noise and tested using entirely different segments of the same nonstationary noise type. The training set was expanded through a noise-perturbation technique. Substantial benefit was observed for hearing-impaired listeners in both noise types, despite the use of unseen noise segments during the operational stage. Interestingly, normal-hearing listeners displayed benefit in babble but not in cafeteria noise. These results highlight the importance of evaluating these algorithms not only in human subjects, but in members of the actual target population. [Work supported by NIH.]